Modélisation de la distribution des espèces en Afrique : état de l'art et perspectives

Abstract

Modeling the distribution of species has become a very popular tool to aid in anticipation and decision-making in the conservation of biological resources. The current review aimed to summarize the existing work in this area in Africa, identify the shortcomes and suggest research perspectives. To this end, online databases were used to identify and download the publications considered here. Of the 229 studies carried out in Africa, 167 published between 2003 and 2019 in 100 journals were deemed relevant for this review. Almost 27 % of the publications were made in Benin and South Africa. Twenty-three methodological modeling approaches have been identified and divided into two main groups: empirical or correlative methods and mechanistic ones. Correlative methods and more specifically the principle of maximum entropy Maxent (113 studies out of 167) were the most used (about 68% of the studies). This could be explained by the fact that the correlative methods are open access, less restrictive, less demanding on data and more easily implementable. Among the weaknesses noted in those studies were the level of data quality, not always sufficient to ensure the validity of the drawn conclusions. Likewise, sampling biases are generally ignored in the specification of models while the observed distribution of occurrences is rather the result of sampling biases or the inaccessibility of certain areas rather than the environmental preferendum of the modeled species. Finally, most studies on the continent did not include species demography or biotic interactions. Exploring such aspects could help to improve the quality of the models.

Keywords

• validation
• bioclimate
• model
• projection
• presence-only data
• sources of bias